EP0258559A2 - Cache memory coherency control provided with a read in progress indicating memory - Google Patents
Cache memory coherency control provided with a read in progress indicating memory Download PDFInfo
- Publication number
- EP0258559A2 EP0258559A2 EP87109195A EP87109195A EP0258559A2 EP 0258559 A2 EP0258559 A2 EP 0258559A2 EP 87109195 A EP87109195 A EP 87109195A EP 87109195 A EP87109195 A EP 87109195A EP 0258559 A2 EP0258559 A2 EP 0258559A2
- Authority
- EP
- European Patent Office
- Prior art keywords
- memory
- cache
- data
- rip
- control means
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/14—Handling requests for interconnection or transfer
- G06F13/16—Handling requests for interconnection or transfer for access to memory bus
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F12/00—Accessing, addressing or allocating within memory systems or architectures
- G06F12/02—Addressing or allocation; Relocation
- G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
- G06F12/0802—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
- G06F12/0806—Multiuser, multiprocessor or multiprocessing cache systems
- G06F12/084—Multiuser, multiprocessor or multiprocessing cache systems with a shared cache
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F12/00—Accessing, addressing or allocating within memory systems or architectures
- G06F12/02—Addressing or allocation; Relocation
- G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
- G06F12/0802—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
- G06F12/0844—Multiple simultaneous or quasi-simultaneous cache accessing
- G06F12/0855—Overlapped cache accessing, e.g. pipeline
- G06F12/0859—Overlapped cache accessing, e.g. pipeline with reload from main memory
Definitions
- the present invention relates to cache memory systems and more particularly to cache memory systems utilized in multiprocessor systems.
- cache memories have been highly effective in increasing the throughput of small and large monoprocessor and multiprocessor systems.
- caches are frequently configured in a so-called private cache arrangement in which the cache memory is dedicated to a single processor.
- systems have increased the number of processing units, each connecting a cache memory which, in turn, connected to main memory and the other units of the overall system via an asynchronous system bus.
- the independently operating processing units produce unsynchronized overlapping requests to main memory. This can substantively affect cache coherency which is the ability of the cache to accurately and correctly track the contents of main memory. The result is that processing units can be forced to operate with stale data which could eventually bring the system to a halt.
- a write through approach employs a listener device which detects the occurrence of any write operations made to main memory and updates the contents of the cache.
- the second approach employs circuits which invalidate the data contents of the cache locations when any other system unit writes into a main memory location which has been mapped into a processing unit's cache.
- one prior art cache system detects when any data is read or written into main memory by another unit prior to the receipt of the requested data which falls within a given address range. If it does, the cache is bypassed and the requested data is transferred to the requesting processing unit. While this ensures cache coherency, the process of bypassing cache can result in decreased system efficiency. This occurs as the system becomes busier due to the addition of more and faster processing units resulting in a substantial decrease in cache hit ratio.
- a cache memory subsystem couples to main memory through interface circuits via an asynchronous system bus in common with a number of central processing subsystems which have similar interface circuits.
- the cache memory subsystem includes multilevel directory memory and buffer memory pipeline stages shareable by a number of processing units.
- read in process (RIP) memory means is operatively associated with one of the pipeline stages.
- the RIP memory means is set to a predetermined state in response to read request from a processing unit which produces a miss condition indicating that the requested data is not resident in the buffer memory stage.
- Setting occurs when listener apparatus coupled to the system bus detects that the cache memory subsystem has forwarded a request for the missing data to main memory and presents that memory request to the cache memory subsystem resulting in the preallocation of directory memory.
- the RIP memory means identifies the address of the directory location within a specific level which was preallocated.
- the listener apparatus maintains the contents of the buffer memory stage coherent by updating its contents in response to any request applied to the system bus by other subsystems.
- the cache memory subsystem prior to the receipt of the requested data from main memory, the cache memory subsystem, in response to detecting that the listener apparatus has been receiving data which would make the buffer memory incoherent, operates to switch control means associated with the RIP memory from a preset state to a different state.
- the directory memory is accessed, the RIP memory means is reset and the latest version of the requested data is forwarded to the requesting processing unit as a function of the state of the control means. That is, when the control means has been switched to a different state, this denotes that the requested data is not the most recent data (e.g. has been modified by another processing unit).
- the latest version of the data stored in the buffer memory is read out and transferred to the requesting processing unit while the data received from main memory is discarded.
- the requested data is written into the buffer memory and transferred to the requesting processing unit.
- the arrangement of the present invention is able to maintain coherency without decreasing the cache efficiency, since requested main memory data will be written into the buffer store unless it proves to be stale (not the most recent). Of course, this presumes that the data meets those requirements necessary for insuring data integrity (e.g. error free).
- the arrangement of the prsent invention ensures that the operation of the cache memory subsystem is properly synchronized with the asynchronous operations of the system bus. That is, the same memory request sent to main memory on the system bus is also presented to the cache memory subsystem during the same cycle resulting in the preallocation of the directory memory. This ensures reliable operation within the pipelined stages.
- the arrangement of the present invention can be used to maintain cache coherency between a pair of the processing units sharing the cache memory subsystem.
- a second memory is associated with the other pipeline and is conditioned to store the same information as the RIP memory.
- the state of the second memory blocks the loading of the other processing unit's data registers causing a miss condition. This allows the other processing unit to stall its operations until such time as the requested main memory data can be provided.
- a single memory associated with the directory stage may be used to maintain coherency between processing units. However, for performance reasons, it is preferable to use a memory in each pipeline stage. This eliminates the need to have to perform read modify write operations which would increase the time of a pipeline stage to perform the required operations. Also, the arrangement allows sufficient time for signalling the processing unit to stall its operations.
- Figure 1 shows a multiprocessor data processing system 10 which includes a plurality of subsystems 14 through 30 which couple in common to a system bus 12.
- the illustrative subsystems include a plurality of central subsystems 14 through 16, a plurality of memory subsystems 20 through 28 and a peripheral subsystem 30.
- Each memory subsystem is organized to include even and odd memory modules.
- An example of such an arrangement is disclosed in U.S. Patent No. 4,432,055.
- Each subsystem includes an interface area which enables the unit or units associated hterewith to transmit or receive requests in the form of commands, interrupts, data or responses/status to another unit on system bus 12 in an asynchronous manner. That is, each interface area can be assumed to include bus interface logic circuits such as those disclosed in U.S. Patent No. 3,995,258, entitled "Data Processing System Having a Data Integrity Technique", invented by George J. Barlow.
- FIG. 14 shows in block diagram form the organization of central subsystem 14.
- Subsystem 14 includes a pair of central processing unit (CPU) subsystems 14-2 and 14-4 coupled to share a cache subsystem 14-6.
- the cache subsystem 14-6 couples to system bus 12 through a first in first out (FIFO) subsystem 14-10 which can be considered as being included within interface area 14-1.
- FIFO first in first out
- each CPU subsystem includes a 32-bit central processing unit (CPU) (i.e., CPU's 14-20 and 14-40), and a virtual memory management unit (VMMU) (i.e., VMMU 14-26 and 14-46) for translating CPU virtual addresses into physical addresses for presentation to cache subsystem 14-6 as part of the memory requests.
- each CPU subsystem includes a read only store (ROS) and a 16-bit ROS data output register (RDR) (i.e., ROS 14-25, 14-44 and RDR 14-25, 14-45).
- ROS read only store
- RDR 16-bit ROS data output register
- each ROS is conditioned to read out a 16-bit microinstruction word into its data output (RDR) register which defines the type of operation to be performed during the cycle (firmware step/box).
- RDR data output
- the clock circuits within each CPU subsystem i.e., circuits 14-22 and 14-42) establish the basic timing for its subsystem under the control or cache subsystem 14-6 as explained herein.
- the elements of each CPU subsystem can be constructed from standard integrated circuit chips.
- cache subsytstem 14-6 is organized into a source address generation section and two separate pipeline stages, each with its own decode and control circuits.
- the source address generation section includes blocks 14-62 and 14-64 which perform the functions of source address selecting and incrementing.
- the first pipeline stage is an address stage and includes the directory and associated memory circuits of blocks 14-66 through 14-76, arranged as shown. This stage performs the functions of latching the generated source address, directory searching and hit comparing.
- the first pipeline stage provides as an output information in the form of a level number and a column address. The opeations of the first pipeline stage are clocked by timing signals generated by the timing and control circuits of block 14-60.
- the information from the first stage is immediately passed onto the second pipeline stage leaving the first stage available for the next source request.
- the second pipeline stage is a data stage and includes the data buffer and associated memory circuits of blocks 14-80 through 14-96, arranged as shown. This stage performs the functions of accessing the requested data from the buffer memories 14-88 and 14-90, or replacing/storing data with data received from FIFO subsystem 14-10.
- the second pipeline stage provides a 36-bit data word for transfer to one of the CPU subsystems.
- the operations of the second pipeline stage are clocked by timing signals generated by the timing and control circuits of block 14-60.
- the different blocks of the first and second pipeline stages are constructed from standard integrated circuits, such as those described in the "The TTL Data Book, Column 3", Copyrighted 1984, by Texas Instruments Inc. and in the "Advanced Micro Devices Programmable Array Logic Handbook", Copyright 1983, by Advanced Micro Devices, Inc.
- the address selector circuit of block 14-62 is constructed from two sets of six 74AS857 multiplexer chips cascaded to select one of four addresses.
- the swap multiplexer of block 14-92 is constructed from the same type chips.
- the latches of blocks 14-68 and 14-72 are constructed from 74AS843 D-type latch chips.
- the swap multiplexer and data register circuits of block 14-70 are constructed from a single clocked programmable array logic element, such as part number AMPA16R6A, manufactured by Advanced Micro Devices, Inc.
- the directory and associated memories 14-74 and 14-76 are constructed from 8-bit slice cache address comparator circuits having aprt number TMS2150JL, manufactured by Texas Instruments Incorporated and a 4K x 4-bit memory chip having part number IMS1421, manufactured by INMOS Corporation.
- the address and data registers 14-80 through 14-84 and 14-94 and 14-96 are constructed from 9-bit interface flip-flops having part number SN74AS823, manufactured by Texas Instruments, Inc.
- the buffer memories 14-88 and 14-90 are also constructed from 4K x 4-bit memory chips having part number IMS1421, manufactured by INMOS Corporation and a 4096 x 1 static RAM chip having part number AM2147, manufactured by Advanced Micro Devices, Inc.
- the address increment circuits of block 14-64 are constructed from standard ALU chips designated by part number 74AS181A and a programmable array logic element having part number AmPAL16L8A, manufactured by Advanced Micro Devices, Inc.
- the first and second levels of command register and decode circuits of block 14-66 and 14-86 utilize clocked programmable array logic elements having part numbers AmPAL16R4A and AmPAL16R6A, manufactured by Advanced Micro Devices, Inc. These circuits generate the required selection, read and write control signals as indicated in Figure 2 (i.e., signals SWAPLT+00, SWAPRT+00, POLDDT-OL, PILDDT-OL, POLDDFOR, F1LDDT-OR). For further details, reference may be made to the equations of the Appendix.
- cache subsystem 14-6 is organized into even and odd sections which permit two data words to be accessed simultaneously in response to either an odd or even memory address.
- cache addressing arrangement See U.S. Patent No. 4,378,591 which is assigned to the same assignee as named herein.
- FIG. 2 also shows in block form, FIFO subsystem 14-10 which includes the FIFO control and clocking circuits of block 14-11 which couples to a replacement address register 14-12 and to system bus 12.
- FIFO subsystem 14-10 receives all of the information transferred between any two subsystems on system bus 12. When the information is for updating data in main memory, the inforamtion is coded to indicate such updating or replacement operation.
- FIFO subsystem 14-10 also receives any new data resulting from a memory request being forwarded to system bus 12 by cache subsystem 14-6. Both update and new data are stored as requests within a buffer memory included within subsystem 14-10.
- FIFO control circuits decode each request and initiate the appropriate cycles of operation which result in address, data and commands being applied to different parts of cache subsystem 14-6 as seen from Figure 2.
- FIFO subsystem can be considered conventional in design and take the form of the type of FIFO circuits disclosed in U.S. Patent No. 4,195,340 which is assigned to the same assignee as named herein.
- timing and control circuits of block 14-60 The basic timing for each of the subsystems of Figure 2 is established by the timing and control circuits of block 14-60. Such control permits the conflict-free sharing of cache subsystem 14-6 by CPU subsystems 14-2 and 14-4 and FIFO subsystem 14-10.
- the circuits of block 14-60 are described in greater detail in the first related patent application. Briefly, these circuits include address select logic circuits which generate control signals for conditioning address selector 14-62 to select one of the subsystems 14-2, 14-4 and 14-10 as a request address source.
- block 14-60 includes pipeline clock circuits of block 14-620 which define the different types of cache memory cycles which can initiate the start of the pipeline resulting in the generation of a predetermined sequence of signals in response to each request. That is, first and second signals, respectively, indicate a cache request for service by CPU0 subsystem 14-2 and CPU1 subsystem 14-4 while other signals indicate cache requests for service by FIFO subsystem 14-10.
- Figure 3 shows the organization of the even and odd directory memory and buffer memory pipeline stages, according to the present invention.
- the directory and associated memory circuits of blocks 14-74 and 14-76 each include a multilevel 4K x 20-bit directory memory 14-740 and a second half bus cycle in process 512 x 4-bit memory (SIP) 14-742.
- SIP 4-bit memory
- the directory memory 14-740 in response to a cache address, generates eight hit output signals (HIT0-7) which are applied to the hit decode circuits of block 14-86.
- the directory memory 14-740 is written in response to write enable signals LVWR0 through LVWR7 from circuits 14-66.
- the SIP memory 14-742 in response to the cache column address, generates four output signals (SIP0-3) which are also applied to the hit decode circuits of block 14-86.
- the SIP memory 14-742 is written with level number and SIP bit signals as a function of input data signals SIPDT0-3 and a write enable signal SIPWRT generated by the circuits of block 14-66. For further details as to how these signals are generated, reference may be made to the equations of the Appendix.
- the hit decode circuits of block 14-86 include the hit decode circuits of block 14-860 and the SIP hit decode circuits of block 14-862.
- separate clocked PLA elements are used to construct the hit decode circuits of blocks 14-860 and 14-862.
- hit decode circuits of block 14-860 In response to the input signals shown, hit decode circuits of block 14-860 generate hit level number output signals HIT0-2 and HIT. These signals indicate the level at which the hit occurred and the hit condition occurrence during a second half bus cycle or update cycle (i.e., normal operation). Simultaneously, the priority encoded SIP hit decode circuits of block 14-862 generate a SIP output signal SIPHIT upon detecting the prsence of a potential coherency problem due to both processing units having requested the same data. For further details as to how these signals are generated, reference may be made to the equations in the Appendix.
- the outputs from the circuits of blocks 14-860 and 14-682 are applied as inputs to the circuits of the data buffer pipeline stage as shown.
- the data buffer and associated emmory circuits of blocks 14-88 and 14-90 each include 4K x 16-bit data buffer 14-880, a 4K x 1-bit read in process (RIP) memory 14-882 and associated D-type clocked control flip-flop 14-884 and AND gate 14-886 all of which connect as shown.
- the RIP memory 14-882 is written as a function of an input data signal RIPDTI and write a enable signal RIPWRT generated by the circuits of block 14-86. For further details as to how these signals are generated, reference may be made to the equations in the Appendix.
- the data ouput signal RIPDAT from the RIP memory 14-882 is loaded into control flip-flop 14-884 via AND gate 14-886 in rsponse to a HIT signal from block 14-860 during an update cycle (i.e., when signal FIUPDATE is a ONE).
- the flip-flop 14-884 is preset to a binary ZERO state during an allocate cycle (i.e., when signal is a binary ZERO).
- the ouput signal RIPFLP from the flop-flop 14-884 is applied as an input to a data selector circuit 14-866 which is also included within block 14-86.
- data selector circuit 14-866 As a function of the state of signal RIPFLP, data selector circuit 14-866 generates output enable signals CADREGEN, CADBWEN and CADBEN which selectively condition data register 14-82 and data buffer memory 14-880 to provide the data signals to be loaded into one of the CPU data registers 14-94 and 14-96. For further details as to how these signals are generated, reference may be made to the equations in the Appendix.
- each processing unit can initiate processing requests to memory.
- the memory subsystem normally operates to queue the memory requests and process each request in sequence.
- one processing unit may be reading a word or a part of a word from the same location another processing unit is updating or writing. Since the processing unit can apply the write request to system bus 12 before the memory subsystem delivers the information requested by the first processing unit, the information actually delivered to the first processing unit will be incoherent. That is, the information will contain a mixture of old and new information.
- the apparatus of the present invention not only enables the cache subsystem to provide coherent data to its associated processing unit or units, but it provides the latest version of the data being requested.
- the read in process RIP memory apparatus is included in both the even and odd cache sections so as to provide coherency for each of the two 16-bit words.
- this apparatus can be expanded to provide the same degree of coherency for each byte of each word. This can be done by duplicating RIP memory apparatus.
- Figure 4a illustrates the operation of a system configured as a single processing unit which couples to system bus 12 through a single cache.
- Figure 3 need only include a single memory (either SIP memory 14-742 or RIP memory 14-882).
- SIP memory 14-742 or RIP memory 14-882 For performance reasons, it is desirable to include the memory in the first pipeline stage. It will be appreciated that both memories 14-742 and 14-882 are equivalent.
- the memory is organized into a 4K by 1-bit memory. Each of the 512 addressable locations has 8 levels associated therewith. Thus, the levels are encoded and their value determined through accessing the memory.
- the output SIPHIT from the SIP hit decode circuits of block 14-862 are connected as an input to AND gate 14-886 in place of the RIPDAT signal.
- the RIP memory 14-882 is therefore eliminated.
- CPU 0 couples to cache subsystem 14-6.
- This signal causes the interface area 14-1 to issue a read request to the main memory subsystem 20-2 requesting the data word stored at location A0. That is, memory read request is applied to the system bus 12 and, upon being accepted by the memory subsystem 20-2 is acknowledged by its interface area 20-1 circuits.
- the same read request is loaded into the FIFO subsystem 14-10 from system bus 12.
- the bus request is synchronized with the cache pipeline timing.
- the cache subsystem preallocates the directory by writing the row address portion of the read request cache address into the location specified by the cache column address at the designated level.
- the level is determined by decoding the level number bits FIDT16-18 or FIDT19-21 from FIFO subsystem 14-10 which forces a corresponding one of the directory write enable signals LVWR0 through LVWR7 to a binary ZERO.
- RIP memory 14-742 is set to a predetermind state for denoting that there is a read in process. That is, the level number and read in process bits are written into the location specified by the cache column address in response to signals SIPDT0-3 and SIPWRT. At that time, signal presets RIP flip-flop 14-884 to a binary ZERO state.
- the FIFO subsystem 14-11 listens for any write request made by another processing unit. The receipt of any such request causes the FIFO subsystem 14-10 to initiate a memory write update cycle (i.e., forces signal FIUPDATE to a binary ONE). In the same manner as described above, the memory write cache address is applied via latches 14-68/72 to the directory memory 14-740 and memory 14-742 of Figure 3.
- the directory memory 14-740 operates to generate hit signals HIT0-7 indicating whether or not the data being updated is stored in data buffer 14-880 and if stored, the level in the data buffer where it is stored.
- memory 14-742 operates to generate level number and RIP bit signals SIP0-3 indicating if a read in process state exists for that column address.
- the hit decode circuits 14-680 operate to convert signals HIT0-7 into a three-bit hit level code which corresponds to signals HIT#0-2, in addition to generating hit output signal HIT indicating the occurrence of a hit condition.
- the SIP hit decode circuits 14-862 operate to compare the level number signals SIP0-2 with the converted signals HIT#0-2 for determining whether or not the data being updated will produce a coherency problem. When there is a match and the RIP bit signal is true, the hit decode circuits 14-862 force signal SIPHIT to a binary ONE indicating the potential coherency problem.
- the occurrence of a directory hit causes the write update data loaded into data register 14-82 to be written into data buffer 14-880, in response to signals CADBWEN and CADBEN (i.e., both binary ZEROS).
- the data word is written into the location specified by cache column address signals COLADO-8 and hit level number signals HIT#0-2 which were loaded into buffer address register 14-80/84, in rsponse to clocking signal PIPEOB+OA.
- the occurrence of a RIP hit causes RIP flip-flop 14-884 to be switched to a binary ONE, in response to signals SIPHIT and FIUPDATE and timing signal PIPEOB-OB.
- the RIP flip-flop When set to a binary ONE, the RIP flip-flop ensures that the cache subsystem furnishes coherent data to the processing unit notwithstanding the fact that the requested data word was updated by another processing unit. More specifically, when the requested data word is received from main memory, during a second half bus cycle, another FIFO cycle is initiated. This results in the requested data word being transferred via the SWAPMUX data register 14-70 to data register 14-82.
- the cache address signals are applied to the directory memory 14-740. Since the directory has been preallocated, reading the directory results in the generation of a hit condition by hit decode circuits 14-680, as indicated in Figure 4a. Next, the state of RIP flip-flop 14-884 is examined. The state of signal RIPFLP conditions data selector 14-866 to select the latest version of the data word for transfer to the processing unit. That is, when signal RIPFLP is a binary ONE as in the present example, this indicates that an update cycle to the same location as the requested data had occurred before the SHBC cycle which returned the requested data word.
- data selector circuits 14-866 are conditioned to discard the returned data word stored in data register 14-82 and enable data buffer 14-880 for reading out the updated data word to the CPU data register 14-94/96. This is accomplished by data selector circuit 14-866 forcing signal DREGEN to a binary ONE disabling data register 14-82 and signal CADBEN to a binary ZERO enabling data buffer 14-880.
- data selector circuit 14-860 enables data register 14-82 to apply the data word to CPU data register 14-94/96 (i.e., forces signal DREGEN to a binary ZERO) while conditioning data buffer 14-880 for writing the data word into location A0 at the specified level (i.e., forces signal CADBWEN to a binary ZERO and signal CADBEN to a binary ONE).
- the RIP bit is reset to a binary ZERO in response to signal SIPWRT.
- data input signal SIPDT3 is a binary ZERO.
- the apparatus of the present inention is able to both maintain coherency and provide a processing unit with the latest version of a coherent pair of data words. That is, the even and odd sections of the cache subsystem include the apparatus of Figure 3 and operate in the manner illustrated in Figure 4a to inhibit the transfer of incoherent data words to CPU 14-2.
- Figure 4b illustrates the operation of the present invention when used in a system configured as cache subsystem 14-6 of Figure 2.
- both of the memories 14-742 and 14-882 are utilized in the different pipeline stages as shown in Figure 3.
- the arrangement provides performance benefits in that the memories in each pipeline stage are either written or read (i.e., no read modify write operations are required) thereby minimizing pipeline delays.
- CPU0 generates a memory read request to read word 0 from location A0 of main memory subsystem 20 which produces a miss condition.
- the interface 14-1 issues a memory read request to main memory subsystem 20 for data word 0 in location A0.
- CPU0 signals the memory subsystem 20 that the read is a double word read (i.e., forces a double word bus line to a binary ONE state). This causes subsystem 20 to furnish data words from the even and odd memory modules (i.e., locations A0 and A0+1) to CPU0.
- the memory read request is loaded into FIFO subsystem 14-10 and results in a FIFO cache cycle request.
- the directory memory 14-740 of the even and odd cache sections is preallocated by writing the read request cache row address into the location A0 specified by the cache column address at the designated level (Xi).
- the SIP memory 14-742 is accessed and the level numbr Xi along with the SIP bit are written into the location A0 specified by the cache column address.
- the RIP memory 14-882 of the even and odd cache sections is accessed.
- the RIP bit location specified by the cache column address hit number signals HIT#0-2 is set to a binary ONE in response to signals RIPDTI and RIPWRT generated by the circuits 14-86.
- the RIP flip-flop 14-884 of each cache section is preset to a binary ZERO in the manner previously described.
- the cache subsystem 14-6 in response to each request determines whether the request is for a CPU read, write update or second half bus cycle. Assuming it is a CPU read cycle, the memory read request generated by CUP1 (i.e., CUP0 awaiting results of memory read request) is examined when it is presented to the cache subsystem 20 by CPU1. That is, the cache address loaded into latches 14-68/72 accesses directory memory 14-740 and SIP memory 14-742 of the even and odd cache sections. When CUP1 is requesting one of the data words previously requested by CPU0, the directory memory 14-740 generates ouput hit signals HIT 0-7 indicating the occurrence of a hit condition at level Xi.
- the SIP hit decode circuits 14-862 operate to force signal SIPHIT to a binary ONE when the SIP bit is a binary ONE and the hit output signals HIT0-7 match the level number signals SIP0-2.
- the SIPHIT signal is used to indicate the presence of a coherency problem between CPU0 and CPU1.
- SIPHIT signal is used to block the loading of CPU1 data registers with data from data buffer 14-880 which produces a miss condition at interface area 14-1. This effectively cancels the request as far as cache subsystem 20 is concerned. To simplify the design, the interface area 14-1 still forwards CPU1's memory request to main memory subsystem 20 in response to the miss condition.
- the memory write request is examined when it is presented to the cache subsystem 20 by FIFO subsystem 14-10 following the receipt of an acknowledgement from system bus 12.
- the cache address is applied via latches 14-68/72 to directory memory 14-740 and when the data word being updated is the same as that requested by CPU0, hit decode circuits 14-860 operate to generate a hit.
- main memory subsystem 20 When main memory subsystem 20 completes its processing of the CPU0 memory read request, it operates to return data words 0 and 1 on system bus 12 during a second half bus cycle. As seen from Figure 4b, a SHBC cycle is initiated during which the returned data words together with the cache address is presented to cache subsystem 20 by FIFO subsystem 14-10.
- the cache address is applied to directory memory 14-740 which results in hit decode circuits 14-860 generating a hit. Also, SIP memory 14-742 is accessed and the SIP bit is reset to a binary ZERO in response to signal SIPWWRT.
- the state of RIP flip-flop 14-884 determines whether data buffer 14-880 or data register 14-82 contains the latest version of the requested data words.
- signal RIPFLP conditions data selector circuit 14-866 to inhibit the transfer of the SHBC data from register 14-82. It also enables the rereading of cache data buffer 14-880 by forcing signal CADBEN to a binary ZERO and the delivery of the updated data words to the CPU0 data register.
- each processing unit within a system or in the case of a shared cache each pair of processing units within a system receive coherent data and the latest version of such coherent data. This is accomplished with the addition of few circuits within one or more of the pipeline stages of the cache subsystem of the preferred embodiment. For performance reasons, memory circuits are included in each pipeline stage of the shared cache subsystem.
Abstract
Description
-
- 1. The patent application of James W. Keeley and Thomas F. Joyce entitled, "Multiprocessor Shared Pipeline Cache Memory", filed on September 27, 1984, bearing serial number 06/655,473, which is assigned to the same assignee as this patent application.
- 2. The patent application of James W. Keeley entitled, "A Coherent Cache System Shared by a Pair of Processing Units", filed on even date as this application, bearing serial number _________, which is assigned to the same assignee as this patent application.
- The present invention relates to cache memory systems and more particularly to cache memory systems utilized in multiprocessor systems.
- It is well known that cache memories have been highly effective in increasing the throughput of small and large monoprocessor and multiprocessor systems. In such systems, caches are frequently configured in a so-called private cache arrangement in which the cache memory is dedicated to a single processor.
- To increase system throughput, systems have increased the number of processing units, each connecting a cache memory which, in turn, connected to main memory and the other units of the overall system via an asynchronous system bus. In such systems, the independently operating processing units produce unsynchronized overlapping requests to main memory. This can substantively affect cache coherency which is the ability of the cache to accurately and correctly track the contents of main memory. The result is that processing units can be forced to operate with stale data which could eventually bring the system to a halt.
- In general, there have been two basic approaches employed in maintaining cache coherency. The first termed a write through approach employs a listener device which detects the occurrence of any write operations made to main memory and updates the contents of the cache. The second approach employs circuits which invalidate the data contents of the cache locations when any other system unit writes into a main memory location which has been mapped into a processing unit's cache.
- Employing the write through approach, one prior art cache system detects when any data is read or written into main memory by another unit prior to the receipt of the requested data which falls within a given address range. If it does, the cache is bypassed and the requested data is transferred to the requesting processing unit. While this ensures cache coherency, the process of bypassing cache can result in decreased system efficiency. This occurs as the system becomes busier due to the addition of more and faster processing units resulting in a substantial decrease in cache hit ratio.
- Accordingly, it is a primary object of the present invention to provide a cache memory which maintains cache coherency without decreasing system efficiency.
- It is a further object of the present invention to maintain a cache coherency notwithstanding unsynchronized overlapping memory requests by a number of independently operated processing units which connect to a system bus through independent interfaces.
- The above objects and advantages are achieved in a preferred embodiment of the present invention. In the preferred embodiment, a cache memory subsystem couples to main memory through interface circuits via an asynchronous system bus in common with a number of central processing subsystems which have similar interface circuits. The cache memory subsystem includes multilevel directory memory and buffer memory pipeline stages shareable by a number of processing units.
- According to the present invention, read in process (RIP) memory means is operatively associated with one of the pipeline stages. The RIP memory means is set to a predetermined state in response to read request from a processing unit which produces a miss condition indicating that the requested data is not resident in the buffer memory stage. Setting occurs when listener apparatus coupled to the system bus detects that the cache memory subsystem has forwarded a request for the missing data to main memory and presents that memory request to the cache memory subsystem resulting in the preallocation of directory memory. When so set, the RIP memory means identifies the address of the directory location within a specific level which was preallocated. The listener apparatus maintains the contents of the buffer memory stage coherent by updating its contents in response to any request applied to the system bus by other subsystems.
- In the preferred embodiment, prior to the receipt of the requested data from main memory, the cache memory subsystem, in response to detecting that the listener apparatus has been receiving data which would make the buffer memory incoherent, operates to switch control means associated with the RIP memory from a preset state to a different state. When the requested data is received, the directory memory is accessed, the RIP memory means is reset and the latest version of the requested data is forwarded to the requesting processing unit as a function of the state of the control means. That is, when the control means has been switched to a different state, this denotes that the requested data is not the most recent data (e.g. has been modified by another processing unit). Therefore, the latest version of the data stored in the buffer memory is read out and transferred to the requesting processing unit while the data received from main memory is discarded. However, when the control means has remained in an initially prset state, this denotes that the data received from main memory is the most recent data. Thus, the requested data is written into the buffer memory and transferred to the requesting processing unit.
- The arrangement of the present invention is able to maintain coherency without decreasing the cache efficiency, since requested main memory data will be written into the buffer store unless it proves to be stale (not the most recent). Of course, this presumes that the data meets those requirements necessary for insuring data integrity (e.g. error free).
- In addition to maintaining coherency, the arrangement of the prsent invention ensures that the operation of the cache memory subsystem is properly synchronized with the asynchronous operations of the system bus. That is, the same memory request sent to main memory on the system bus is also presented to the cache memory subsystem during the same cycle resulting in the preallocation of the directory memory. This ensures reliable operation within the pipelined stages.
- In addition to the above, the arrangement of the present invention can be used to maintain cache coherency between a pair of the processing units sharing the cache memory subsystem. In such circumstances, a second memory is associated with the other pipeline and is conditioned to store the same information as the RIP memory. When the other processing unit makes a memory read request for the same data, the state of the second memory blocks the loading of the other processing unit's data registers causing a miss condition. This allows the other processing unit to stall its operations until such time as the requested main memory data can be provided.
- A single memory associated with the directory stage may be used to maintain coherency between processing units. However, for performance reasons, it is preferable to use a memory in each pipeline stage. This eliminates the need to have to perform read modify write operations which would increase the time of a pipeline stage to perform the required operations. Also, the arrangement allows sufficient time for signalling the processing unit to stall its operations.
- The novel features which are believed to be characteristic of the invention both as to its organization and method of operation, together with further objects and advantages will be better understood from the following description when considered in connection with the accompanying drawings. It is to be expressly understood, however, that each of the drawings is given for the purpose of illustration and description only and is not intended as a definition of the limits of the present invention.
-
- Figure 1 is a block diagram of a system which includes the apparatus of the present invention.
- Figure 2 is a block diagram of one of the central subsystems of Figure 1 constructed according to the present inention.
- Figure 3 shows in greater detail the circuits of the pipeline stages of Figure 2.
- Figures 4a and 4b are flow diagrams used to explain the operation of the apparatus of the present invention used in differently configured central subsystems.
- Figure 1 shows a multiprocessor
data processing system 10 which includes a plurality ofsubsystems 14 through 30 which couple in common to asystem bus 12. The illustrative subsystems include a plurality ofcentral subsystems 14 through 16, a plurality ofmemory subsystems 20 through 28 and aperipheral subsystem 30. Each memory subsystem is organized to include even and odd memory modules. An example of such an arrangement is disclosed in U.S. Patent No. 4,432,055. - Each subsystem includes an interface area which enables the unit or units associated hterewith to transmit or receive requests in the form of commands, interrupts, data or responses/status to another unit on
system bus 12 in an asynchronous manner. That is, each interface area can be assumed to include bus interface logic circuits such as those disclosed in U.S. Patent No. 3,995,258, entitled "Data Processing System Having a Data Integrity Technique", invented by George J. Barlow. - The organization of each of the
central subsystems 14 through 16 is the same. Figure 2 shows in block diagram form the organization ofcentral subsystem 14.Subsystem 14 includes a pair of central processing unit (CPU) subsystems 14-2 and 14-4 coupled to share a cache subsystem 14-6. The cache subsystem 14-6 couples tosystem bus 12 through a first in first out (FIFO) subsystem 14-10 which can be considered as being included within interface area 14-1. - As seen from Figure 2, both CPU subsystems 14-2 and 14-4 are identical in construction. That is, each CPU subsystem includes a 32-bit central processing unit (CPU) (i.e., CPU's 14-20 and 14-40), and a virtual memory management unit (VMMU) (i.e., VMMU 14-26 and 14-46) for translating CPU virtual addresses into physical addresses for presentation to cache subsystem 14-6 as part of the memory requests. Also, each CPU subsystem includes a read only store (ROS) and a 16-bit ROS data output register (RDR) (i.e., ROS 14-25, 14-44 and RDR 14-25, 14-45).
- At the beginning of each cycle, each ROS is conditioned to read out a 16-bit microinstruction word into its data output (RDR) register which defines the type of operation to be performed during the cycle (firmware step/box). The clock circuits within each CPU subsystem (i.e., circuits 14-22 and 14-42) establish the basic timing for its subsystem under the control or cache subsystem 14-6 as explained herein. The elements of each CPU subsystem can be constructed from standard integrated circuit chips.
- As seen from Figure 2, cache subsytstem 14-6 is organized into a source address generation section and two separate pipeline stages, each with its own decode and control circuits. The source address generation section includes blocks 14-62 and 14-64 which perform the functions of source address selecting and incrementing. The first pipeline stage is an address stage and includes the directory and associated memory circuits of blocks 14-66 through 14-76, arranged as shown. This stage performs the functions of latching the generated source address, directory searching and hit comparing. The first pipeline stage provides as an output information in the form of a level number and a column address. The opeations of the first pipeline stage are clocked by timing signals generated by the timing and control circuits of block 14-60.
- The information from the first stage is immediately passed onto the second pipeline stage leaving the first stage available for the next source request. The second pipeline stage is a data stage and includes the data buffer and associated memory circuits of blocks 14-80 through 14-96, arranged as shown. This stage performs the functions of accessing the requested data from the buffer memories 14-88 and 14-90, or replacing/storing data with data received from FIFO subsystem 14-10. Thus, the second pipeline stage provides a 36-bit data word for transfer to one of the CPU subsystems. Again, the operations of the second pipeline stage are clocked by timing signals generated by the timing and control circuits of block 14-60.
- The different blocks of the first and second pipeline stages are constructed from standard integrated circuits, such as those described in the "The TTL Data Book,
Column 3", Copyrighted 1984, by Texas Instruments Inc. and in the "Advanced Micro Devices Programmable Array Logic Handbook", Copyright 1983, by Advanced Micro Devices, Inc. For example, the address selector circuit of block 14-62 is constructed from two sets of six 74AS857 multiplexer chips cascaded to select one of four addresses. The swap multiplexer of block 14-92 is constructed from the same type chips. The latches of blocks 14-68 and 14-72 are constructed from 74AS843 D-type latch chips. The swap multiplexer and data register circuits of block 14-70 are constructed from a single clocked programmable array logic element, such as part number AMPA16R6A, manufactured by Advanced Micro Devices, Inc. - The directory and associated memories 14-74 and 14-76, shown in greater detail in Figure 3, are constructed from 8-bit slice cache address comparator circuits having aprt number TMS2150JL, manufactured by Texas Instruments Incorporated and a 4K x 4-bit memory chip having part number IMS1421, manufactured by INMOS Corporation. The address and data registers 14-80 through 14-84 and 14-94 and 14-96 are constructed from 9-bit interface flip-flops having part number SN74AS823, manufactured by Texas Instruments, Inc.
- The buffer memories 14-88 and 14-90, shown in greater detail in Figure 3, are also constructed from 4K x 4-bit memory chips having part number IMS1421, manufactured by INMOS Corporation and a 4096 x 1 static RAM chip having part number AM2147, manufactured by Advanced Micro Devices, Inc. The address increment circuits of block 14-64 are constructed from standard ALU chips designated by part number 74AS181A and a programmable array logic element having part number AmPAL16L8A, manufactured by Advanced Micro Devices, Inc.
- The first and second levels of command register and decode circuits of block 14-66 and 14-86, respectively, utilize clocked programmable array logic elements having part numbers AmPAL16R4A and AmPAL16R6A, manufactured by Advanced Micro Devices, Inc. These circuits generate the required selection, read and write control signals as indicated in Figure 2 (i.e., signals SWAPLT+00, SWAPRT+00, POLDDT-OL, PILDDT-OL, POLDDFOR, F1LDDT-OR). For further details, reference may be made to the equations of the Appendix.
- As seen from Figure 2, cache subsystem 14-6 is organized into even and odd sections which permit two data words to be accessed simultaneously in response to either an odd or even memory address. For further information about this type of cache addressing arrangement, reference may be made to U.S. Patent No. 4,378,591 which is assigned to the same assignee as named herein.
- Figure 2 also shows in block form, FIFO subsystem 14-10 which includes the FIFO control and clocking circuits of block 14-11 which couples to a replacement address register 14-12 and to
system bus 12. FIFO subsystem 14-10 receives all of the information transferred between any two subsystems onsystem bus 12. When the information is for updating data in main memory, the inforamtion is coded to indicate such updating or replacement operation. FIFO subsystem 14-10 also receives any new data resulting from a memory request being forwarded tosystem bus 12 by cache subsystem 14-6. Both update and new data are stored as requests within a buffer memory included within subsystem 14-10. FIFO control circuits decode each request and initiate the appropriate cycles of operation which result in address, data and commands being applied to different parts of cache subsystem 14-6 as seen from Figure 2. For the purpose of the present invention, FIFO subsystem can be considered conventional in design and take the form of the type of FIFO circuits disclosed in U.S. Patent No. 4,195,340 which is assigned to the same assignee as named herein. - The basic timing for each of the subsystems of Figure 2 is established by the timing and control circuits of block 14-60. Such control permits the conflict-free sharing of cache subsystem 14-6 by CPU subsystems 14-2 and 14-4 and FIFO subsystem 14-10. The circuits of block 14-60 are described in greater detail in the first related patent application. Briefly, these circuits include address select logic circuits which generate control signals for conditioning address selector 14-62 to select one of the subsystems 14-2, 14-4 and 14-10 as a request address source.
- Also, block 14-60 includes pipeline clock circuits of block 14-620 which define the different types of cache memory cycles which can initiate the start of the pipeline resulting in the generation of a predetermined sequence of signals in response to each request. That is, first and second signals, respectively, indicate a cache request for service by CPU0 subsystem 14-2 and CPU1 subsystem 14-4 while other signals indicate cache requests for service by FIFO subsystem 14-10.
- These requests can be summarized as follows:
- 1. CPU0 READ CYCLE A CPU0 read occurs in response to a cache request initiated by ROS 14-24 during a first time slot/interval when
CPU port 0 within interface 14-1 is not busy. The address supplied by CPU0 subsystem 14-2 is furnished to the first pipeline stage and the directory is read. When a hit is detected, indicating that the requested data is stored in the data buffer, the buffer is read and the data is clocked into the CPU0 data register. When a miss is detected, the CPU0 port is made busy, the request is forwarded to memory to fetch the requested data. - 2. CPU1 READ CYCLE A CPU1 read occurs in response to a cache request initiated by ROS 14-44 during a third time slot/interval when
CPU port 1 within interface 14-1 is not busy. - 3. SECOND HALF BUS CYCLE A second half bus cycle occurs in response to a first type of cache request initiated by FIFO subsystem 14-10 for data requested from either main memory or an I/O device being returned on
system bus 12 during a first or third time slot/interval when FIFO subsystem 14-10 has a request stored. when FIFO subsystem 14-10 furnishes data from an I/O device to the first pipeline state, it passes therethrough without changing the states of any memories and is clocked into the appropriate CPU data register. Data from main memory is written into the cache data buffers and is clocked into the appropriate CPU data registers. - 4. MEMORY WRITE UPDATE CYCLE A memory write update cycle occurs in response to a second type of cache request initiated by FIFO subsystem 14-10 for replacement or update data received from
system bus 12, upon acknowledgement of such data during a first or third time slot/interval when FIFO subsystem 14-10 has a request stored. FIFO subsystem 14-10 furnishes data to the first pipeline stage resulting in the reading of the directory memory. When a hit is detected, the replacement data is written into the buffer memory. - 5. FIFO ALLOCATION CYCLE A FIFO allocation occurs in response to a CPU0 or CPU1 READ CYCLE which results in a miss being detected. The CPU port is made busy and the request is forwarded to memory to fetch the requested data. Upon the memory read request being acknowledged, the CPU read request is loaded into the FIFO subsystem registers and control circuits included in the subsystem initiate a request for a FIFO cycle of operation (i.e., force signal CYFIFO=1). Signals specifying the type of request and level number information are applied as inputs to the command register and decode circuits of block 14-66. These signals include FIMREF (memory reference), (memory read) and FIDT16-18/19-21 (level number). The signals FIMREF and initiate a FIFO allocation cycle (i.e., FIALOCYC=1).
- Figure 3 shows the organization of the even and odd directory memory and buffer memory pipeline stages, according to the present invention. As seen from Figure 3, the directory and associated memory circuits of blocks 14-74 and 14-76, each include a multilevel 4K x 20-bit directory memory 14-740 and a second half bus cycle in process 512 x 4-bit memory (SIP) 14-742.
- The directory memory 14-740, in response to a cache address, generates eight hit output signals (HIT0-7) which are applied to the hit decode circuits of block 14-86. The directory memory 14-740 is written in response to write enable signals LVWR0 through LVWR7 from circuits 14-66. The SIP memory 14-742, in response to the cache column address, generates four output signals (SIP0-3) which are also applied to the hit decode circuits of block 14-86. The SIP memory 14-742 is written with level number and SIP bit signals as a function of input data signals SIPDT0-3 and a write enable signal SIPWRT generated by the circuits of block 14-66. For further details as to how these signals are generated, reference may be made to the equations of the Appendix.
- The hit decode circuits of block 14-86 include the hit decode circuits of block 14-860 and the SIP hit decode circuits of block 14-862. In the preferred embodiment, separate clocked PLA elements are used to construct the hit decode circuits of blocks 14-860 and 14-862.
- In response to the input signals shown, hit decode circuits of block 14-860 generate hit level number output signals HIT0-2 and HIT. These signals indicate the level at which the hit occurred and the hit condition occurrence during a second half bus cycle or update cycle (i.e., normal operation). Simultaneously, the priority encoded SIP hit decode circuits of block 14-862 generate a SIP output signal SIPHIT upon detecting the prsence of a potential coherency problem due to both processing units having requested the same data. For further details as to how these signals are generated, reference may be made to the equations in the Appendix.
- The outputs from the circuits of blocks 14-860 and 14-682 are applied as inputs to the circuits of the data buffer pipeline stage as shown. Accoridng to the preferred embodiment of the prsent invention, the data buffer and associated emmory circuits of blocks 14-88 and 14-90, each include 4K x 16-bit data buffer 14-880, a 4K x 1-bit read in process (RIP) memory 14-882 and associated D-type clocked control flip-flop 14-884 and AND gate 14-886 all of which connect as shown. The RIP memory 14-882 is written as a function of an input data signal RIPDTI and write a enable signal RIPWRT generated by the circuits of block 14-86. For further details as to how these signals are generated, reference may be made to the equations in the Appendix.
- The data ouput signal RIPDAT from the RIP memory 14-882 is loaded into control flip-flop 14-884 via AND gate 14-886 in rsponse to a HIT signal from block 14-860 during an update cycle (i.e., when signal FIUPDATE is a ONE). The flip-flop 14-884 is preset to a binary ZERO state during an allocate cycle (i.e., when signal is a binary ZERO). The ouput signal RIPFLP from the flop-flop 14-884 is applied as an input to a data selector circuit 14-866 which is also included within block 14-86.
- As a function of the state of signal RIPFLP, data selector circuit 14-866 generates output enable signals CADREGEN, CADBWEN and CADBEN which selectively condition data register 14-82 and data buffer memory 14-880 to provide the data signals to be loaded into one of the CPU data registers 14-94 and 14-96. For further details as to how these signals are generated, reference may be made to the equations in the Appendix.
- With reference to the flow diagrams of Figures 4a and 4b, the operation of the apparatus of the present invention shown in Figure 3 will now be described. In the system of Figure 1, each processing unit can initiate processing requests to memory. The memory subsystem normally operates to queue the memory requests and process each request in sequence. In certain instances, one processing unit may be reading a word or a part of a word from the same location another processing unit is updating or writing. Since the processing unit can apply the write request to
system bus 12 before the memory subsystem delivers the information requested by the first processing unit, the information actually delivered to the first processing unit will be incoherent. That is, the information will contain a mixture of old and new information. - While it may be possible to avoid incoherency by requiring the imposition of locking instructions for processing units sharing common areas of memory, this may not always be practical. Also, it may be difficult to detect and prevent violations.
- The apparatus of the present invention not only enables the cache subsystem to provide coherent data to its associated processing unit or units, but it provides the latest version of the data being requested. In the preferred embodiment, the read in process RIP memory apparatus is included in both the even and odd cache sections so as to provide coherency for each of the two 16-bit words. Of course, this apparatus can be expanded to provide the same degree of coherency for each byte of each word. This can be done by duplicating RIP memory apparatus.
- The sequence of opeations shown in Figure 4a illustrate the operation of a system configured as a single processing unit which couples to
system bus 12 through a single cache. In this case, Figure 3 need only include a single memory (either SIP memory 14-742 or RIP memory 14-882). For performance reasons, it is desirable to include the memory in the first pipeline stage. It will be appreciated that both memories 14-742 and 14-882 are equivalent. To increase the speed of operation of RIP memory 14-882, the memory is organized into a 4K by 1-bit memory. Each of the 512 addressable locations has 8 levels associated therewith. Thus, the levels are encoded and their value determined through accessing the memory. - In the case of the single processing unit configuration, the output SIPHIT from the SIP hit decode circuits of block 14-862 are connected as an input to AND gate 14-886 in place of the RIPDAT signal. The RIP memory 14-882 is therefore eliminated.
- With reference to the flow chart of Figure 4a, it will be assumed that
CPU 0 couples to cache subsystem 14-6. When CPU 14-2 requests thedata word 0 stored at location A0 of the even memory module of memory subsystem 20-2 of Figure, the cache subsystem 14-6 hit decode circuits of block 14-660 signal a miss condition (i.e., signal HIT=0). This signal causes the interface area 14-1 to issue a read request to the main memory subsystem 20-2 requesting the data word stored at location A0. That is, memory read request is applied to thesystem bus 12 and, upon being accepted by the memory subsystem 20-2 is acknowledged by its interface area 20-1 circuits. - Upon being acknowledged, the same read request is loaded into the FIFO subsystem 14-10 from
system bus 12. In this manner, the bus request is synchronized with the cache pipeline timing. The cache memory address is applied via address selector circuit 14-62 to the address latches 14-68/74 while certain command and data bits are applied to the command register and decode circuits of block 14-66. This occurs whenever there is a free pipeline stage and FIFO subsystem 14-10 initiated a request for a FIFO cycle (i.e., signal CYFIFO=1). - The command bits specifying the type of FIFO request are decoded by the circuits of block 14-66. Since the request is a memory read request, signals FIMREF and are binary ONES. These signals force signal FIUPDATE to a binary ONE initiating a FIFO allocation cycle of operation (i.e., signal FIALOCYC=1).
- As seen from Figure 4a, the cache subsystem preallocates the directory by writing the row address portion of the read request cache address into the location specified by the cache column address at the designated level. The level is determined by decoding the level number bits FIDT16-18 or FIDT19-21 from FIFO subsystem 14-10 which forces a corresponding one of the directory write enable signals LVWR0 through LVWR7 to a binary ZERO.
- At the same time, RIP memory 14-742 is set to a predetermind state for denoting that there is a read in process. That is, the level number and read in process bits are written into the location specified by the cache column address in response to signals SIPDT0-3 and SIPWRT. At that time, signal presets RIP flip-flop 14-884 to a binary ZERO state.
- As seen from Figure 4a, the FIFO subsystem 14-11 listens for any write request made by another processing unit. The receipt of any such request causes the FIFO subsystem 14-10 to initiate a memory write update cycle (i.e., forces signal FIUPDATE to a binary ONE). In the same manner as described above, the memory write cache address is applied via latches 14-68/72 to the directory memory 14-740 and memory 14-742 of Figure 3.
- During each write update cycle, the directory and associated memory circuits are accessed by cache column address signals COLADO-8. The directory memory 14-740 operates to generate hit signals HIT0-7 indicating whether or not the data being updated is stored in data buffer 14-880 and if stored, the level in the data buffer where it is stored. At the same time, memory 14-742 operates to generate level number and RIP bit signals SIP0-3 indicating if a read in process state exists for that column address.
- The hit decode circuits 14-680 operate to convert signals HIT0-7 into a three-bit hit level code which corresponds to signals HIT#0-2, in addition to generating hit output signal HIT indicating the occurrence of a hit condition. The SIP hit decode circuits 14-862 operate to compare the level number signals SIP0-2 with the converted signals HIT#0-2 for determining whether or not the data being updated will produce a coherency problem. When there is a match and the RIP bit signal is true, the hit decode circuits 14-862 force signal SIPHIT to a binary ONE indicating the potential coherency problem.
- As seen from Figure 4a, the occurrence of a directory hit causes the write update data loaded into data register 14-82 to be written into data buffer 14-880, in response to signals CADBWEN and CADBEN (i.e., both binary ZEROS). The data word is written into the location specified by cache column address signals COLADO-8 and hit level number signals HIT#0-2 which were loaded into buffer address register 14-80/84, in rsponse to clocking signal PIPEOB+OA. At this time, the occurrence of a RIP hit causes RIP flip-flop 14-884 to be switched to a binary ONE, in response to signals SIPHIT and FIUPDATE and timing signal PIPEOB-OB.
- When set to a binary ONE, the RIP flip-flop ensures that the cache subsystem furnishes coherent data to the processing unit notwithstanding the fact that the requested data word was updated by another processing unit. More specifically, when the requested data word is received from main memory, during a second half bus cycle, another FIFO cycle is initiated. This results in the requested data word being transferred via the SWAPMUX data register 14-70 to data register 14-82.
- Also, the cache address signals are applied to the directory memory 14-740. Since the directory has been preallocated, reading the directory results in the generation of a hit condition by hit decode circuits 14-680, as indicated in Figure 4a. Next, the state of RIP flip-flop 14-884 is examined. The state of signal RIPFLP conditions data selector 14-866 to select the latest version of the data word for transfer to the processing unit. That is, when signal RIPFLP is a binary ONE as in the present example, this indicates that an update cycle to the same location as the requested data had occurred before the SHBC cycle which returned the requested data word. Therefore, data selector circuits 14-866 are conditioned to discard the returned data word stored in data register 14-82 and enable data buffer 14-880 for reading out the updated data word to the CPU data register 14-94/96. This is accomplished by data selector circuit 14-866 forcing signal DREGEN to a binary ONE disabling data register 14-82 and signal CADBEN to a binary ZERO enabling data buffer 14-880.
- As seen from Figure 4a, when RIP flip-flop 14-886 has not been set to a binary ONE, the data word is written into data buffer 14-880 and transferred to the
CPU 0. That is, data selector circuit 14-860 enables data register 14-82 to apply the data word to CPU data register 14-94/96 (i.e., forces signal DREGEN to a binary ZERO) while conditioning data buffer 14-880 for writing the data word into location A0 at the specified level (i.e., forces signal CADBWEN to a binary ZERO and signal CADBEN to a binary ONE). - Also, during the SHBC cycle, the RIP bit is reset to a binary ZERO in response to signal SIPWRT. At that time, data input signal SIPDT3 is a binary ZERO.
- From the above, it is seen that the apparatus of the present inention is able to both maintain coherency and provide a processing unit with the latest version of a coherent pair of data words. That is, the even and odd sections of the cache subsystem include the apparatus of Figure 3 and operate in the manner illustrated in Figure 4a to inhibit the transfer of incoherent data words to CPU 14-2.
- Figure 4b illustrates the operation of the present invention when used in a system configured as cache subsystem 14-6 of Figure 2. To ensure coherency between the two processing units CPU0 and CPU1 which share the cache subsystem, both of the memories 14-742 and 14-882 are utilized in the different pipeline stages as shown in Figure 3. The arrangement provides performance benefits in that the memories in each pipeline stage are either written or read (i.e., no read modify write operations are required) thereby minimizing pipeline delays.
- As seen from Figure 4b, it is assumed that CPU0 generates a memory read request to read
word 0 from location A0 ofmain memory subsystem 20 which produces a miss condition. In the same manner as described above, the interface 14-1 issues a memory read request tomain memory subsystem 20 fordata word 0 in location A0. Also, CPU0 signals thememory subsystem 20 that the read is a double word read (i.e., forces a double word bus line to a binary ONE state). This causessubsystem 20 to furnish data words from the even and odd memory modules (i.e., locations A0 and A0+1) to CPU0. - As shown in Figure 4b, the memory read request is loaded into FIFO subsystem 14-10 and results in a FIFO cache cycle request. As previously discussed, the directory memory 14-740 of the even and odd cache sections is preallocated by writing the read request cache row address into the location A0 specified by the cache column address at the designated level (Xi). At the same time, the SIP memory 14-742 is accessed and the level numbr Xi along with the SIP bit are written into the location A0 specified by the cache column address.
- Next, the RIP memory 14-882 of the even and odd cache sections is accessed. The RIP bit location specified by the cache column address hit number signals HIT#0-2 is set to a binary ONE in response to signals RIPDTI and RIPWRT generated by the circuits 14-86. Also, the RIP flip-flop 14-884 of each cache section is preset to a binary ZERO in the manner previously described.
- As seen from Figure 4b, the cache subsystem 14-6 in response to each request determines whether the request is for a CPU read, write update or second half bus cycle. Assuming it is a CPU read cycle, the memory read request generated by CUP1 (i.e., CUP0 awaiting results of memory read request) is examined when it is presented to the
cache subsystem 20 by CPU1. That is, the cache address loaded into latches 14-68/72 accesses directory memory 14-740 and SIP memory 14-742 of the even and odd cache sections. When CUP1 is requesting one of the data words previously requested by CPU0, the directory memory 14-740 generates ouput hit signals HIT 0-7 indicating the occurrence of a hit condition at level Xi. - The SIP hit decode circuits 14-862 operate to force signal SIPHIT to a binary ONE when the SIP bit is a binary ONE and the hit output signals HIT0-7 match the level number signals SIP0-2. The SIPHIT signal is used to indicate the presence of a coherency problem between CPU0 and CPU1.
- As indicated in Figure 4b, SIPHIT signal is used to block the loading of CPU1 data registers with data from data buffer 14-880 which produces a miss condition at interface area 14-1. This effectively cancels the request as far as
cache subsystem 20 is concerned. To simplify the design, the interface area 14-1 still forwards CPU1's memory request tomain memory subsystem 20 in response to the miss condition. - In response to each write update cycle, the memory write request is examined when it is presented to the
cache subsystem 20 by FIFO subsystem 14-10 following the receipt of an acknowledgement fromsystem bus 12. As described above, the cache address is applied via latches 14-68/72 to directory memory 14-740 and when the data word being updated is the same as that requested by CPU0, hit decode circuits 14-860 operate to generate a hit. - This results in the updated data word being written into data buffer 14-880 at the location specified by the cache column address and hit level signals HIT#0-2. Also, at that time, RIP memory 14-882 is accessed. When the RIP bit has been set to a binary ONE signalling that a read is in process, this forces signal RIPDAT to a binary ONE. This causes RIP flip-flop 14-884 to be switched to a binary ONE in response to signals HIT, FIUPDATE and PIPEOB-OB as shown in Figure 4b.
- When
main memory subsystem 20 completes its processing of the CPU0 memory read request, it operates to returndata words system bus 12 during a second half bus cycle. As seen from Figure 4b, a SHBC cycle is initiated during which the returned data words together with the cache address is presented tocache subsystem 20 by FIFO subsystem 14-10. - As described above, the cache address is applied to directory memory 14-740 which results in hit decode circuits 14-860 generating a hit. Also, SIP memory 14-742 is accessed and the SIP bit is reset to a binary ZERO in response to signal SIPWWRT.
- As seen from Figure 4b, the state of RIP flip-flop 14-884 determines whether data buffer 14-880 or data register 14-82 contains the latest version of the requested data words. When set to a binary ONE, signal RIPFLP conditions data selector circuit 14-866 to inhibit the transfer of the SHBC data from register 14-82. It also enables the rereading of cache data buffer 14-880 by forcing signal CADBEN to a binary ZERO and the delivery of the updated data words to the CPU0 data register.
- When RIP flip-flop 14-884 is a binary ZERO, this indicates that the requested data words have not been updated. Hence, the SHBC data words stored in register 14-82 are written into data buffer 14-880 and delivered to the CPU0 data register. As seen from Figure 4b, during the SHBC cycle, the RIP bit accessed from RIP memory 14-882 is reset to a binary ZERO in response to signals RIPDTI and RIPWRT. Since this is not an allocation cycle, signal RIPDTI is a binary ZERO.
- From the above, it is seen how the apparatus of the present invention ensures that each processing unit within a system or in the case of a shared cache, each pair of processing units within a system receive coherent data and the latest version of such coherent data. This is accomplished with the addition of few circuits within one or more of the pipeline stages of the cache subsystem of the preferred embodiment. For performance reasons, memory circuits are included in each pipeline stage of the shared cache subsystem.
-
-
- 1. DBWDRD = Double word read command defined by ROS data bit 4 = 1 and ROS data bit 5 = 0 generated by the decode circuits of block 14-66 which is clocked with signal PIPE0A+0A.
- 2. CPUNUM = CPU number (CPU0 or CPU1) signal generated by the circuits of block 14-66 which is clocked with signal PIPE0A+0A.
- 3. CPUCYL = CPU cycle signal generated by the circuits of block 14-66 and which is clocked with signal PIPE0A+0A.
- 4. EVNHIT = HIT signal generated by the hit decode circuits 14-680 associated with the even directory memory 14-76.
- 5. CMAD22 = Cache
memory address bit 22, generated at the output of selector 14-62. - 6. CMAD22 = Cache
memory address bit 23, generated at the output of selector 14-62, specifies which half (left or right) of data register 14-94 or 14-96 is to be loaded with a data word. - 7. FIAD17 = FIFO address bit 17 from FIFO subsystem 14-10 defines which CPU is to receive the replacement data.
- 8. FIDBWD = FIFO double-wide word command bit from FIFO subsystem 14-11 specifies when the data being returned has two words.
- 9. FISHBA = FIFO second half bus cycle acknowledge signal from 14-11 specifies that the FIFO subsystem requires a cache cycle to process data received from an I/O device or memory during a second half bus cycle SHBC.
- 10. ODDHIT = HIT signal generated by the hit decode circuits 14-680 associated with the odd directory memory 14-74.
- 11. RPMREF = Memory reference signal provided by RAR 14-12 which permits any exception conditions to be taken into account.
- 12. RPAD22 = Replacement address bit 22 from RAR 14-12.
- 13. FIDT16-18/19-21 = The even/odd data bits defining the cache level provided by the FIFO subsystem 14-10.
- 14. CYFIFO = A cycle signal generated by the FIFO cycle select logic circuits of block 14-60 during a free pipeline stage.
- 15. FISHBC = The second half bus cycle signal from FIFO subsystem 14-10.
- 16. WRTPLS = The write pulse signal generated by the circuits of block 14-60 which occurs midway between either clocking signals PIPE0A+0A AND PIPE0A+OB or clocking signals PIPE0B+0A and PIPE0B+0B.
- 17. FIMREF = The bus memory reference signal BSMREF from FIFO subsystem 14-10.
- 18. FIWRIT = The bus memory write signal BSWRIT from FIFO subsystem 14-10.
- It will be appreciated by those skilled in the art that many changes may be made to the preferred embodiment of the present invention. As mentioned, there need only be a single memory associated with the directory memory stage in the case of a single processing unit. Where there is more than one processing unit sharing the cache subsystem, one memory is used for detecting incoherency between processing units while the second memory is used to indicate a read in process state included for increased performance. However, it still may be possible to combine both within a single memory, as well as the function of the RIP control flip-flop with attendant decrease in pipeline stage speed.
- While in accordance with the provisions and statutes there has been illustrated and described the best form of the invention, certain changes may be made without departing from the spirit of the invention as set forth in the appended claims and that in some cases, certain features of the invention may be used to advantage without a corresponding use of other foatures.
Claims (28)
a first stage including directory store means organized into a plurality of levels containing groups of storage locations each location for storing said first address portion of a memory read request generated by said processing unit associated therewith and each different group of locations within said directory store levels being defined by a different one of said second address portions;
a second stage including data store means organized into the same number of levels of said groups of locations as in said directory store means and each different grop of locations within said data store levels being accessed by a different one of said second address portions;
read in process (RIP) memory means included in one of said first and second cache stages, said RIP memory means including a plurality of locations, each location being accessed by a different one of said second address portions;
decode and control means coupled to said directory store means, to said RIP memory means and to said data store means, said decode and control means being operative during a cache allocation cycle in response to each request received from said processing unit for data not stored in said data store means to generate signals for placing the location specified by said second address portion in said RIP memory means in a predetermined state for identifying the data store location which has been preallocated; and
control means coupled to said RIP memory means and to said decode and control means, said control means being ocnditioned by said decode and control means during a cache update cycle to switch to said predetermined state when said RIP memory means signals that a portion of the contents of said data store location which has been preallocated is being updated to maintain coherency prior to the receipt of the requested data and
said control means being operative to condition said data store means for transferring the most recent version of said requested data to said processing unit.
an even directory memory having a plurality of locations for storing a plurality of even addresses; and
an odd directory memory having a plurality of locations for storing a plurality of odd addresses; and wherein
said buffer memory means includes:
an even buffere memory having a plurality of storage locations associated with a different one of a plurality of even addresses; and
an odd buffer memory having a plurality of storage locations associated with a different one of a plurality of odd addresses; and
said RIP memory means includes:
an even RIP memory; and
an odd RIP memory associated with said even directory emmory and said odd directory memory respectively when included in said first cache stage and said even buffer memory and said odd buffer memory respectively when included in said second cache stage.
a first stage including a directory store organized into a plurality of levels containing groups of storage locations, each location for storing said first address portion of a memory read request generated by one of said number of processing units associated therewith and each different group of locations within said directory store levels being defined by a different one of said second address portions;
a second cache stage including:
a data store organized into the same number of levels of said groups of locations as in said directory store and each different group of locations within said data store levels being defined by a different one of said second address portions;
input data register means coupled to said bus, to said data store and to said processing units for receiving data to be written into said data store and transferred to said processing units in response to said requests; and
read in process (RIP) memory means including a plurality of locations, each location defined by a different one of said second address portions;
decode and control means copled to said directory store, to said RIP memory means, to said data store and to said input data register means, said decode and control means being operative during an allocation cycle in response to each request received from said one of said number of processing units for data not stored in said data store to generate signals for placing the location in said RIP memory means specified by certain ones of said signals and said second address portion in a predetermined state for identifying the data store location which has been preallocated; and
control means coupled to said RIP memory means and to said decode and control means, said control means being conditioned by said decode and control means during an update cycle to switch to said predetermined state when said RIP memory means signals that a portion of the data contents of said preallocated data store location is being updated prior to the receipt of the requested data to be stored in the same data location and
said control means being operative to selectively condition said input data register means and said data store during a second half bus cycle for transferring the most recent coherent version of said requested data to said one of said processing units.
said decode and control means being coupled to said SIP memory means, said decode and control means being operative during said allocation cycle in response to said each request received from one of said number of processing units for data not stored in said data store to generate signals for placing the location in said SIP memory means specified by said second address portion in said predetermined state for identifying said data store location which has been preallocated and
said decode and control means being operative in response to the contents of said location read out during a subsequent allocation cycle generated in response to a memory request from another one of said processing units to inhibit the transfer of data to said another processing unit from said preallocated location in cache store thereby maintaining data coherency between said processing units.
an even directory memory having a plurality of locations for storing a plurality of even addresses; and
an odd directory memory having a plurality of locations for storing a plurality of odd addresses; and wherein
said buffer memory means includes:
an even buffer memory having a plurality of storage locations associated with a different one of a plurality of even addresses; and
an odd buffer memory having a plurality of storage locations associated with a different one of a plurality of odd addresses;
said SIP memory means includes:
an even SIP memory; and
an odd SIP memory associated with said even directory memory and said odd directory memory respectively; and
said RIP memory means includes:
an even RIP memory; and
an odd RIP memory associated with said even buffer memory and said odd buffer memory respectively.
a pipelined cache memory subsystem coupled to each of said processing units for receiving said data requests, said cache subsystem comprising:
input selection means for selecting a request address from one of said processing units subsystems during an allocated time slot interval;
a first pipeline cache stage coupled to said input selection means, said pipeline stage including a directory store organized into a plurality of levels containing groups of storage locations, each location for storing said first address portion of a memory read request generated by one of said processing units during said allocated time slot interval and each different group of locations within said directory store levels being accessed by a different one of said second address portions;
a second cache pipeline stage including a data store organized into the same number of levels of said grops of locations as in said directory store and each different group of locations within said data store levels being accessible by a different one of said second address portions during a succeeding time slot interval for transfer of the data contents to the requesting one of said processing units;
read in process (RIP) memory means included in one of said first and second cache stages, said RIP memory means including a plurality of locations, each location being accessed by a different one of said second address portions;
decode and control means coupled to said directory store, to said RIP memory means, and to said data store, said decode and control means being operative during a cache allocation cycle in response to each request received from one of said processing units for data not stored in said data store to gnerate signals for placing the location specified by said second address portion in said RIP memory means in a predetermined state for identifying the data store location which has been preallocated; and
control means coupled to said RIP memory means and to said decode and control means, said control means being conditioned by said decode and control means during a cache update cycle corresponding to an unused allocated time slot interval to switch to said predetermind state when said RIP memory means signals that a portion of the contents of said data store location which has been preallocated is being updated to maintain coherency prior to the receipt of the requested data to be stored in the same data location and
said control means being operative to selectively condition said data store for transferring the most recent version of said requested data to said requesting processing unit.
said decode and control means being coupled to said SIP memory means, said decode and control means being operative during said allocation cycle in response to said each request received from one of said plurality of processing units for data not stored in said data store to generate signals for placing the location in said SIP memory means specified by said second address portion in said predetermind state for identifying said data store location which has been preallocated and
said decode and control means being operative in response to the contents of said location read out during a subsequent allocation cycle generated in response to a memory request from another one of said processing units to inhibit the transfer of data to said another processing unit from said preallocated location in cache store thereby maintaining data coherency between said plurality of processing units.
an even directory memory having a plurality of locations for storing a plurality of even addresses; and
an odd directory memory having a plurality of locations for storing a plurality of odd addresses; and wherein
said buffer memory means includes:
an even buffer memory having a plurality of storage locations associated with a different one of a plurality of even addreses; and
an odd buffer memory having a plurality of storage locations associated with a different one of a plurality of odd addresses;
said SIP memory means includes:
an even SIP memory; and
an odd SIP memory associated with said even directory memory and said odd directory memory respectively; and
said RIP memory means includes:
an even RIP memory; and
an odd RIP memory associated with said even buffer memory and said odd buffer memory respectively.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US879856 | 1986-06-27 | ||
US06/879,856 US4768148A (en) | 1986-06-27 | 1986-06-27 | Read in process memory apparatus |
Publications (3)
Publication Number | Publication Date |
---|---|
EP0258559A2 true EP0258559A2 (en) | 1988-03-09 |
EP0258559A3 EP0258559A3 (en) | 1990-05-23 |
EP0258559B1 EP0258559B1 (en) | 1994-06-22 |
Family
ID=25375022
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP87109195A Expired - Lifetime EP0258559B1 (en) | 1986-06-27 | 1987-06-26 | Cache memory coherency control provided with a read in progress indicating memory |
Country Status (6)
Country | Link |
---|---|
US (1) | US4768148A (en) |
EP (1) | EP0258559B1 (en) |
KR (2) | KR920010916B1 (en) |
AU (1) | AU599671B2 (en) |
CA (1) | CA1284389C (en) |
DE (1) | DE3750107T2 (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0463573A2 (en) * | 1990-06-28 | 1992-01-02 | Bull HN Information Systems Inc. | Efficient error detection in a VLSI central processing unit |
EP0490526A2 (en) * | 1990-12-14 | 1992-06-17 | International Business Machines Corporation | Multi-system data sharing combination and method of operating same |
GB2325541A (en) * | 1997-04-14 | 1998-11-25 | Ibm | Layering cache and architectural-specific functions |
GB2325542A (en) * | 1997-04-14 | 1998-11-25 | Ibm | Layering cache and architectural-specific functions to expedite multiple designs |
GB2325764A (en) * | 1997-04-14 | 1998-12-02 | Ibm | Layering cache and architectural-specific functions to permit generic interface definition |
US6061755A (en) * | 1997-04-14 | 2000-05-09 | International Business Machines Corporation | Method of layering cache and architectural specific functions to promote operation symmetry |
EP1006448A1 (en) * | 1998-12-04 | 2000-06-07 | STMicroelectronics, Inc. | Pipelined non-blocking two level cache system with inherent transaction collision-avoidance |
US20170132153A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for managing cache line updates for writes, reads, and destages in storage write cache |
Families Citing this family (44)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0668735B2 (en) * | 1987-02-09 | 1994-08-31 | 日本電気アイシーマイコンシステム株式会社 | Cache memory |
US4833601A (en) * | 1987-05-28 | 1989-05-23 | Bull Hn Information Systems Inc. | Cache resiliency in processing a variety of address faults |
US5291581A (en) * | 1987-07-01 | 1994-03-01 | Digital Equipment Corporation | Apparatus and method for synchronization of access to main memory signal groups in a multiprocessor data processing system |
US4969117A (en) * | 1988-05-16 | 1990-11-06 | Ardent Computer Corporation | Chaining and hazard apparatus and method |
US4935849A (en) * | 1988-05-16 | 1990-06-19 | Stardent Computer, Inc. | Chaining and hazard apparatus and method |
DE68924306T2 (en) * | 1988-06-27 | 1996-05-09 | Digital Equipment Corp | Multi-processor computer arrays with shared memory and private caches. |
US5101497A (en) * | 1988-09-09 | 1992-03-31 | Compaq Computer Corporation | Programmable interrupt controller |
US5163142A (en) * | 1988-10-28 | 1992-11-10 | Hewlett-Packard Company | Efficient cache write technique through deferred tag modification |
US4980819A (en) * | 1988-12-19 | 1990-12-25 | Bull Hn Information Systems Inc. | Mechanism for automatically updating multiple unit register file memories in successive cycles for a pipelined processing system |
US5125083A (en) * | 1989-02-03 | 1992-06-23 | Digital Equipment Corporation | Method and apparatus for resolving a variable number of potential memory access conflicts in a pipelined computer system |
US5222224A (en) * | 1989-02-03 | 1993-06-22 | Digital Equipment Corporation | Scheme for insuring data consistency between a plurality of cache memories and the main memory in a multi-processor system |
JPH0740247B2 (en) * | 1989-06-20 | 1995-05-01 | 松下電器産業株式会社 | Cache memory device |
EP0491697B1 (en) * | 1989-09-11 | 1997-10-29 | Wang Laboratories, Inc. | Apparatus and method for maintaining cache/main memory consistency |
US5012408A (en) * | 1990-03-15 | 1991-04-30 | Digital Equipment Corporation | Memory array addressing system for computer systems with multiple memory arrays |
JP2822588B2 (en) * | 1990-04-30 | 1998-11-11 | 日本電気株式会社 | Cache memory device |
US5611070A (en) * | 1990-05-10 | 1997-03-11 | Heidelberger; Philip | Methods and apparatus for performing a write/load cache protocol |
US5249284A (en) * | 1990-06-04 | 1993-09-28 | Ncr Corporation | Method and system for maintaining data coherency between main and cache memories |
US5404482A (en) * | 1990-06-29 | 1995-04-04 | Digital Equipment Corporation | Processor and method for preventing access to a locked memory block by recording a lock in a content addressable memory with outstanding cache fills |
US5404483A (en) * | 1990-06-29 | 1995-04-04 | Digital Equipment Corporation | Processor and method for delaying the processing of cache coherency transactions during outstanding cache fills |
US5313609A (en) * | 1991-05-23 | 1994-05-17 | International Business Machines Corporation | Optimum write-back strategy for directory-based cache coherence protocols |
US5353426A (en) * | 1992-04-29 | 1994-10-04 | Sun Microsystems, Inc. | Cache miss buffer adapted to satisfy read requests to portions of a cache fill in progress without waiting for the cache fill to complete |
US5821940A (en) * | 1992-08-03 | 1998-10-13 | Ball Corporation | Computer graphics vertex index cache system for polygons |
US5598551A (en) * | 1993-07-16 | 1997-01-28 | Unisys Corporation | Cache invalidation sequence system utilizing odd and even invalidation queues with shorter invalidation cycles |
US5530933A (en) * | 1994-02-24 | 1996-06-25 | Hewlett-Packard Company | Multiprocessor system for maintaining cache coherency by checking the coherency in the order of the transactions being issued on the bus |
US6076150A (en) * | 1995-08-10 | 2000-06-13 | Lsi Logic Corporation | Cache controller with improved instruction and data forwarding during refill operation |
US6134632A (en) * | 1998-01-26 | 2000-10-17 | Intel Corporation | Controller that supports data merging utilizing a slice addressable memory array |
US6212616B1 (en) * | 1998-03-23 | 2001-04-03 | International Business Machines Corporation | Even/odd cache directory mechanism |
US6618048B1 (en) | 1999-10-28 | 2003-09-09 | Nintendo Co., Ltd. | 3D graphics rendering system for performing Z value clamping in near-Z range to maximize scene resolution of visually important Z components |
US7119813B1 (en) | 2000-06-02 | 2006-10-10 | Nintendo Co., Ltd. | Variable bit field encoding |
US6867781B1 (en) | 2000-08-23 | 2005-03-15 | Nintendo Co., Ltd. | Graphics pipeline token synchronization |
US6811489B1 (en) | 2000-08-23 | 2004-11-02 | Nintendo Co., Ltd. | Controller interface for a graphics system |
US7002591B1 (en) | 2000-08-23 | 2006-02-21 | Nintendo Co., Ltd. | Method and apparatus for interleaved processing of direct and indirect texture coordinates in a graphics system |
US7538772B1 (en) | 2000-08-23 | 2009-05-26 | Nintendo Co., Ltd. | Graphics processing system with enhanced memory controller |
US6700586B1 (en) | 2000-08-23 | 2004-03-02 | Nintendo Co., Ltd. | Low cost graphics with stitching processing hardware support for skeletal animation |
US7576748B2 (en) * | 2000-11-28 | 2009-08-18 | Nintendo Co. Ltd. | Graphics system with embedded frame butter having reconfigurable pixel formats |
US7061502B1 (en) | 2000-08-23 | 2006-06-13 | Nintendo Co., Ltd. | Method and apparatus for providing logical combination of N alpha operations within a graphics system |
US6980218B1 (en) | 2000-08-23 | 2005-12-27 | Nintendo Co., Ltd. | Method and apparatus for efficient generation of texture coordinate displacements for implementing emboss-style bump mapping in a graphics rendering system |
US6636214B1 (en) | 2000-08-23 | 2003-10-21 | Nintendo Co., Ltd. | Method and apparatus for dynamically reconfiguring the order of hidden surface processing based on rendering mode |
US7034828B1 (en) | 2000-08-23 | 2006-04-25 | Nintendo Co., Ltd. | Recirculating shade tree blender for a graphics system |
US6937245B1 (en) | 2000-08-23 | 2005-08-30 | Nintendo Co., Ltd. | Graphics system with embedded frame buffer having reconfigurable pixel formats |
US7184059B1 (en) | 2000-08-23 | 2007-02-27 | Nintendo Co., Ltd. | Graphics system with copy out conversions between embedded frame buffer and main memory |
US6825851B1 (en) | 2000-08-23 | 2004-11-30 | Nintendo Co., Ltd. | Method and apparatus for environment-mapped bump-mapping in a graphics system |
US6707458B1 (en) | 2000-08-23 | 2004-03-16 | Nintendo Co., Ltd. | Method and apparatus for texture tiling in a graphics system |
KR100872414B1 (en) * | 2007-08-07 | 2008-12-08 | 정근석 | Hot-water circulayion device |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3723976A (en) * | 1972-01-20 | 1973-03-27 | Ibm | Memory system with logical and real addressing |
US3771137A (en) * | 1971-09-10 | 1973-11-06 | Ibm | Memory control in a multipurpose system utilizing a broadcast |
US4494190A (en) * | 1982-05-12 | 1985-01-15 | Honeywell Information Systems Inc. | FIFO buffer to cache memory |
DE3439302A1 (en) * | 1983-10-27 | 1985-05-09 | Hitachi, Ltd., Tokio/Tokyo | STORAGE CONTROL DEVICE |
EP0159712A2 (en) * | 1984-04-27 | 1985-10-30 | Bull HN Information Systems Inc. | Control means in a digital computer |
EP0165823A2 (en) * | 1984-06-22 | 1985-12-27 | Fujitsu Limited | Tag control circuit for buffer storage |
EP0176972A2 (en) * | 1984-09-27 | 1986-04-09 | Bull HN Information Systems Inc. | Multiprocessor shared pipeline cache memory |
Family Cites Families (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4084234A (en) * | 1977-02-17 | 1978-04-11 | Honeywell Information Systems Inc. | Cache write capacity |
US4156906A (en) * | 1977-11-22 | 1979-05-29 | Honeywell Information Systems Inc. | Buffer store including control apparatus which facilitates the concurrent processing of a plurality of commands |
US4161024A (en) * | 1977-12-22 | 1979-07-10 | Honeywell Information Systems Inc. | Private cache-to-CPU interface in a bus oriented data processing system |
US4245304A (en) * | 1978-12-11 | 1981-01-13 | Honeywell Information Systems Inc. | Cache arrangement utilizing a split cycle mode of operation |
US4314331A (en) * | 1978-12-11 | 1982-02-02 | Honeywell Information Systems Inc. | Cache unit information replacement apparatus |
US4349874A (en) * | 1980-04-15 | 1982-09-14 | Honeywell Information Systems Inc. | Buffer system for supply procedure words to a central processor unit |
US4415970A (en) * | 1980-11-14 | 1983-11-15 | Sperry Corporation | Cache/disk subsystem with load equalization |
US4445174A (en) * | 1981-03-31 | 1984-04-24 | International Business Machines Corporation | Multiprocessing system including a shared cache |
JPS58102381A (en) * | 1981-12-15 | 1983-06-17 | Nec Corp | Buffer memory |
US4472774A (en) * | 1982-09-27 | 1984-09-18 | Data General Corp. | Encachement apparatus |
-
1986
- 1986-06-27 US US06/879,856 patent/US4768148A/en not_active Expired - Lifetime
-
1987
- 1987-06-26 EP EP87109195A patent/EP0258559B1/en not_active Expired - Lifetime
- 1987-06-26 CA CA000540643A patent/CA1284389C/en not_active Expired - Lifetime
- 1987-06-26 DE DE3750107T patent/DE3750107T2/en not_active Expired - Fee Related
- 1987-06-26 AU AU74786/87A patent/AU599671B2/en not_active Ceased
- 1987-06-27 KR KR1019870006638A patent/KR920010916B1/en not_active IP Right Cessation
- 1987-06-27 KR KR1019870006637A patent/KR920008430B1/en not_active IP Right Cessation
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3771137A (en) * | 1971-09-10 | 1973-11-06 | Ibm | Memory control in a multipurpose system utilizing a broadcast |
US3723976A (en) * | 1972-01-20 | 1973-03-27 | Ibm | Memory system with logical and real addressing |
US4494190A (en) * | 1982-05-12 | 1985-01-15 | Honeywell Information Systems Inc. | FIFO buffer to cache memory |
DE3439302A1 (en) * | 1983-10-27 | 1985-05-09 | Hitachi, Ltd., Tokio/Tokyo | STORAGE CONTROL DEVICE |
US4683533A (en) * | 1983-10-27 | 1987-07-28 | Hitachi, Ltd. | Storage control system using plural buffer address arrays |
EP0159712A2 (en) * | 1984-04-27 | 1985-10-30 | Bull HN Information Systems Inc. | Control means in a digital computer |
EP0165823A2 (en) * | 1984-06-22 | 1985-12-27 | Fujitsu Limited | Tag control circuit for buffer storage |
EP0176972A2 (en) * | 1984-09-27 | 1986-04-09 | Bull HN Information Systems Inc. | Multiprocessor shared pipeline cache memory |
Cited By (40)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0463573A2 (en) * | 1990-06-28 | 1992-01-02 | Bull HN Information Systems Inc. | Efficient error detection in a VLSI central processing unit |
EP0463573A3 (en) * | 1990-06-28 | 1993-02-03 | Bull Hn Information Systems Inc. | Efficient error detection in a vlsi central processing unit |
EP0490526A2 (en) * | 1990-12-14 | 1992-06-17 | International Business Machines Corporation | Multi-system data sharing combination and method of operating same |
EP0490526A3 (en) * | 1990-12-14 | 1993-11-10 | Ibm | Multi-system data sharing combination and method of operating same |
US6032226A (en) * | 1997-04-14 | 2000-02-29 | International Business Machines Corporation | Method and apparatus for layering cache and architectural specific functions to expedite multiple design |
GB2325542A (en) * | 1997-04-14 | 1998-11-25 | Ibm | Layering cache and architectural-specific functions to expedite multiple designs |
GB2325764A (en) * | 1997-04-14 | 1998-12-02 | Ibm | Layering cache and architectural-specific functions to permit generic interface definition |
US5937172A (en) * | 1997-04-14 | 1999-08-10 | International Business Machines Corporation | Apparatus and method of layering cache and architectural specific functions to permit generic interface definition |
GB2325541A (en) * | 1997-04-14 | 1998-11-25 | Ibm | Layering cache and architectural-specific functions |
US6061755A (en) * | 1997-04-14 | 2000-05-09 | International Business Machines Corporation | Method of layering cache and architectural specific functions to promote operation symmetry |
US6061762A (en) * | 1997-04-14 | 2000-05-09 | International Business Machines Corporation | Apparatus and method for separately layering cache and architectural specific functions in different operational controllers |
GB2325764B (en) * | 1997-04-14 | 2001-12-12 | Ibm | Layering cache and architectural specific functions to permit generic interface definition |
GB2325542B (en) * | 1997-04-14 | 2002-04-03 | Ibm | Layering cache and architectural specific functions to expedite multiple designs |
SG87764A1 (en) * | 1997-04-14 | 2002-04-16 | Ibm | Methods of layering cache and architectural specific functions to expediate multiple designs |
GB2325541B (en) * | 1997-04-14 | 2002-04-17 | Ibm | Layering cache and architectural specific functions |
EP1006448A1 (en) * | 1998-12-04 | 2000-06-07 | STMicroelectronics, Inc. | Pipelined non-blocking two level cache system with inherent transaction collision-avoidance |
US6519682B2 (en) | 1998-12-04 | 2003-02-11 | Stmicroelectronics, Inc. | Pipelined non-blocking level two cache system with inherent transaction collision-avoidance |
US20170132151A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for destage operations from storage write cache |
US9864695B2 (en) * | 2015-11-09 | 2018-01-09 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for managing cache destage rates and thresholds for storage write cache |
US20170131909A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management with cache line manipulation |
US20170132153A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for managing cache line updates for writes, reads, and destages in storage write cache |
US20170132143A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for merging data with existing data on fast writes to storage write cache |
US20170132142A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for writes to storage write cache |
US20170132137A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for reads from storage write cache |
US20170132138A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for managing cache line updates for purges from storage write cache |
US20170132154A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for managing cache destage rates and thresholds for storage write cache |
US20170132152A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for identification of data age in storage write cache |
US20170132145A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for reads with partial read hits from storage write cache |
US20170132146A1 (en) * | 2015-11-09 | 2017-05-11 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for simultaneous read and destage operations from storage write cache |
US9940250B2 (en) * | 2015-11-09 | 2018-04-10 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for writes to storage write cache |
US9940258B2 (en) * | 2015-11-09 | 2018-04-10 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for merging data with existing data on fast writes to storage write cache |
US9940253B2 (en) * | 2015-11-09 | 2018-04-10 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for destage operations from storage write cache |
US9940251B2 (en) * | 2015-11-09 | 2018-04-10 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for reads from storage write cache |
US9940252B2 (en) * | 2015-11-09 | 2018-04-10 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for reads with partial read hits from storage write cache |
US9940254B2 (en) * | 2015-11-09 | 2018-04-10 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for simultaneous read and destage operations from storage write cache |
US9940256B2 (en) * | 2015-11-09 | 2018-04-10 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for managing cache line updates for writes, reads, and destages in storage write cache |
US9940249B2 (en) * | 2015-11-09 | 2018-04-10 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management with cache line manipulation |
US9940255B2 (en) * | 2015-11-09 | 2018-04-10 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for identification of data age in storage write cache |
US9940257B2 (en) * | 2015-11-09 | 2018-04-10 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for managing cache line updates for purges from storage write cache |
US10078595B2 (en) | 2015-11-09 | 2018-09-18 | International Business Machines Corporation | Implementing hardware accelerator for storage write cache management for managing cache destage rates and thresholds for storage write cache |
Also Published As
Publication number | Publication date |
---|---|
DE3750107D1 (en) | 1994-07-28 |
CA1284389C (en) | 1991-05-21 |
EP0258559B1 (en) | 1994-06-22 |
AU599671B2 (en) | 1990-07-26 |
KR880000858A (en) | 1988-03-30 |
KR920008430B1 (en) | 1992-09-28 |
KR920010916B1 (en) | 1992-12-24 |
AU7478687A (en) | 1988-01-07 |
KR880000861A (en) | 1988-03-30 |
DE3750107T2 (en) | 1995-02-23 |
US4768148A (en) | 1988-08-30 |
EP0258559A3 (en) | 1990-05-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US4768148A (en) | Read in process memory apparatus | |
US4785395A (en) | Multiprocessor coherent cache system including two level shared cache with separately allocated processor storage locations and inter-level duplicate entry replacement | |
EP0176972B1 (en) | Multiprocessor shared pipeline cache memory | |
US4833601A (en) | Cache resiliency in processing a variety of address faults | |
KR100190351B1 (en) | Apparatus and method for reducing interference in two-level cache memory | |
US4995041A (en) | Write back buffer with error correcting capabilities | |
US5809280A (en) | Adaptive ahead FIFO with LRU replacement | |
US5522069A (en) | Symmetric multiprocessing system with unified environment and distributed system functions | |
EP0232526B1 (en) | Paged virtual cache system | |
US5590379A (en) | Method and apparatus for cache memory access with separate fetch and store queues | |
US4586133A (en) | Multilevel controller for a cache memory interface in a multiprocessing system | |
EP0283628A2 (en) | Bus interface circuit for digital data processor | |
EP0303648B1 (en) | Central processor unit for digital data processing system including cache management mechanism | |
EP0732656B1 (en) | Cache MMU system | |
EP1067459B1 (en) | Symmetric multiprocessing system with unified environment and distributed system functions | |
GB2037466A (en) | Computer with cache memory | |
EP0302926B1 (en) | Control signal generation circuit for arithmetic and logic unit for digital processor | |
EP0206841A2 (en) | Microprocessor having separate instruction and data interfaces | |
EP0418220B1 (en) | Destination control logic for arithmetic and logic unit for digital data processor | |
EP0121372A2 (en) | Hierarchy of control stores for overlapped data transmission | |
IE901529A1 (en) | Write back buffer with error correcting capabilities |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
AK | Designated contracting states |
Kind code of ref document: A2 Designated state(s): DE FR GB IT |
|
PUAL | Search report despatched |
Free format text: ORIGINAL CODE: 0009013 |
|
AK | Designated contracting states |
Kind code of ref document: A3 Designated state(s): DE FR GB IT |
|
17P | Request for examination filed |
Effective date: 19901106 |
|
17Q | First examination report despatched |
Effective date: 19920421 |
|
RAP1 | Party data changed (applicant data changed or rights of an application transferred) |
Owner name: BULL HN INFORMATION SYSTEMS INC. |
|
GRAA | (expected) grant |
Free format text: ORIGINAL CODE: 0009210 |
|
AK | Designated contracting states |
Kind code of ref document: B1 Designated state(s): DE FR GB IT |
|
ITF | It: translation for a ep patent filed |
Owner name: BARZANO' E ZANARDO ROMA S.P.A. |
|
REF | Corresponds to: |
Ref document number: 3750107 Country of ref document: DE Date of ref document: 19940728 |
|
ET | Fr: translation filed | ||
RAP2 | Party data changed (patent owner data changed or rights of a patent transferred) |
Owner name: BULL HN INFORMATION SYSTEMS INC. |
|
PLBE | No opposition filed within time limit |
Free format text: ORIGINAL CODE: 0009261 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: FR Effective date: 19950428 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: FR Payment date: 19950503 Year of fee payment: 9 |
|
26N | No opposition filed | ||
REG | Reference to a national code |
Ref country code: FR Ref legal event code: ST |
|
REG | Reference to a national code |
Ref country code: GB Ref legal event code: IF02 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: DE Payment date: 20040526 Year of fee payment: 18 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: IT Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES;WARNING: LAPSES OF ITALIAN PATENTS WITH EFFECTIVE DATE BEFORE 2007 MAY HAVE OCCURRED AT ANY TIME BEFORE 2007. THE CORRECT EFFECTIVE DATE MAY BE DIFFERENT FROM THE ONE RECORDED. Effective date: 20050626 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: DE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20060103 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: GB Payment date: 20060817 Year of fee payment: 20 |
|
REG | Reference to a national code |
Ref country code: GB Ref legal event code: PE20 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: GB Free format text: LAPSE BECAUSE OF EXPIRATION OF PROTECTION Effective date: 20070625 |